http://ocw.mit.edu
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
b
(1)
f f()dx
)
= F(b) - F(a). = F(x)
(The last equality just gives another way of writing F(b) - F(a)that is in widespread use.)
Still another way of writing the theorem is to observe that F(x) is an antiderivative for f(x),
or as it-is sometimes called, an indefinite integral for f(x); using the standard notation for
indefinite integral and the bracket notation given above, the theorem would be written
f (x) dx =
(1')
f (x)d
Writing the theorem this way makes it look sort of catchy, and more importantly, it avoids
having to introduce the new symbol F(x) for the antiderivative.
In contrast with the above theorem, which every calculus student knows, the Second
Fundamental Theorem is more obscure and seems less useful. The purpose of this chapter is
to explain it, show its use and importance, and to show how the two theorems are related.
To start things off, here it is.
Second Fundamental Theorem.
(2)
f (t)dt;
Set F() =
ISimmons calls this function A(x) on p. 207 (2nd edition); this section of Simmons is another presentation
of much of the material given here.
and a= 0.
Solution. Here we can integrate explicitly by finding an antiderivative (using the first
fundamental theorem):
2tsintdt
F(W) =
- cos t2
-cosx
+ 1;
differentiating by the chain rule, we verify that indeed F'(x) = 2zsinx 2, as predicted by
(2).
O
Ssint
dt.
Find F'(7r/2).
Solution. Neither integration techniques nor integral tables will produce an explicit
antiderivative for the function in the integrand. So we cannot use (1). But we can use (2),
which says that
sin 7r/2
1
F'(R/2) = i/2
/2
Many students feel the Second Fundamental Theorem is "obvious"; these students are
confusing it with the similar-looking
(2')
Let F(x) =
f f(x) dx;
derivative for f(x). But the theorem isn't so useful if you can't find an antiderivative. Most
calculus students think for example that e'- has no antiderivative - "the integral isn't
in the tables", "it can't be integrated" are some of the ways they express this. Even for a
simple function like 11- , it is not obvious what the antiderivative is. Perhaps it doesn't
have any?
The Second Fundamental Theorem provides the answer; it says:
Every contiziuous function f(x) has an antoiderivative:
f (t)dt.
The antiderivative may not be expressible in terms of elementary functions - this is the
difficulty with e- "2 - but it always exists.
Solution. This doesn't look so easy to do explicitly. But the Second Fundamental Theorem
says the following function is an antiderivative:
(3) ..
dt
F(z).-
Discussion. You may feel that this doesn't represent progress: the formula for the
antiderivative is useless. But that's not so: the function F(z) can be calculated by numerical
integration. It can be programmed into a calculator so that when you press an z-value, the
screen will display the corresponding value of F(z) to 12 decimal digits. Pressing another
button will draw the graph of F(z) over any interval on the z-axis that you specify.
Repeating what we said earlier, the integral in (3) should be carefully distinguished from
F(z) =
(f+
1)3/8-4(v
+1)1'
(You can prove this is correct by differentiating it; the 8/3 is put in to ensure that F(O) = 0,
as definition (3)requires.)
The above way of writing F(z) is different from (3). Whether or not it is a better way
depends on what you want to know about F(.) and what use you want to make of it. For
instance,
One important use of the Second Fundamental Theorem is to define new functions. Cal
culus can then be used to study their properties.
To illustrate, we consider first an old function: Inz.Let's pretend we know nothing of
logarithms. We do know that
no-1 .
SXd = n+l'
However, we know no explicit formula for an antiderivative of 1/x, i.e., when n = -1.
We therefore use the Second Fundamental Theorem to define an antiderivative of 1/z,
namely
L(x) =
(5)
(We use 1 as the lower limit of integration since the integrand is not defined at 0.) What
are the properties of this function?
dt
Properties of L(x) =
L-1. L(x) is defined for x > 0 (since 1/t is continuous for t > 0);
L-2. L(1) = 0;
L-5. L(x) is increasing for all z > 0, since L'(x) > 0; its graph is concave (i.e., concave
(6)
L(ab)
a d =t
- = IT + ab dt
Comparing with Property L-6, we see we have to show the last integral on the right above
has the value L(b). To see this, make the change of variable t = au and apply the change
of'variable rule (see (7), p. PI.2 in these notes). You get successively
t= au,
dt
t
t
dt = adu,
adu
au
atu
do
du
*u
We have to change the limits on the integral also: t = a and t = ab correspond respectively
to u = 1,u = b. Thus the rule for changing variable in a definite integral gives
jabdt
=
du
Once we have this, the other properties of the logarithm follow in a standard way.
L-7. L(1/a) = -L(a), since L(1/a) + L(a) = L(! - a) = L(1) = 0 .
L-8. lim L(z) = oo; namely, L(z) is incrieasing and L(2") increases without bound as
n--+oo, since L(2") = nL(2), by Property 6; note that L(2) > 0 since f(z) is increasing. 0
In our definition of L(x), the number e appears as the unique number such that
L(e) =
1.
Such a number exists by the Intermediate Value Theorem, 2 since L(z) is increasing, contin
uous (since it has a derivative), and gets bigger than 1.
We now turn to a second example of using the Second Fundamental Theorem to define
a function F(x) - this time, the function will be genuinely new. It is closely related to
an important function in probability and statistics, the error function erf x. (Statistics
oriented calculators have a button for it.) The two functions differ only by a change of scale
on the x- and y-axes. There is no simpler or more elementary expression for F(x).
Example 5. Define a function F(z) by F(z) =
e-' dt.
Sketch the graph of F(x), indicating relative maxima and minima', points of inflection,
saAmmeatrie.
P(1) ro 1 C&hlJ
RatimQ+a
e-t
-.
-.
.-
-.
..
..
x
..
F'(x) = e-
by the Second Fundamental Theorem; since the exponential
positive, this shows F(z) is increasing for all x, and therefore it has no relative ,
minima.
F"(z) = -2xe- 2 ; since F"(z) < 0 for z > 0, the graph of F(x) is
concave (down) when x > 0. Similarly, it is convex (concave up) for z < 0,
and it h
a nint
po
of inflection at x = 0
W-
-=_r
-- Jo
=J
F(., A.
),I
=j.
--
flt~ fIt
Jof
'&
-+.... -WM.
roughly comparable to the area of the trapezoid shown, which about .7.
Simmons, p. 78
-x
'4
-X I
(9)
f(t) dt
a f(s)Az,
since the area of the vertical strip under the curve is approximately the same as the area of
the rectangle. Dividing, we have
AF
where the error in the approximation is bounded by the height of the small curved triangle.
Since .f(t)is continuous, the error is small compared with f(x), and disappears when we
pass to the limit as Ax -- 0; we get therefore
F'(x) = lim
= f(x).
Note that if f(t) were discontinuous at the point x, the result would not be true; as the
picture shows, the approximation in (9) would not be true.
First Fundamental Theorem: Proof4
We want to show that if f(z) is continuous and f(x)'= F'(x), then
f(x) dz = F(b) - F(a).
(10)
We begin by defining
(11)
(z)=
f (t) dt;
G'(x) = f (x),
3
A
4
then
Since G'(x) = f(x) = F'(x), we have (G(x) - F(s))' = 0, which implies G(x) - F(x) = C,
i.e.,
(12)
G(x) = F(x) + C,
C = -- F(a).
Finally, put x = b in equation (12), and use the above value for C:
Remark. Both fundamental theorems say that differentiation and definite integration
are inverse operations: each undoes what the other does. In the First Fundamental Theorem
we differentiate, then integrate:
F(x) --
F'(x)
f)
-W
dA
f(t)
---
d
f(t) dt = f(x) .
In both cases, the theorem says that we end up essentially where we started - only "essen
tially" because of the additive constant in F(x).
(Of course, differentiation and indefinite integration are also inverse operations, but this
is trivial - it's just a restating of the definition of indefinite integration.)
Exercises: Section 3D